On Wed, 21 Aug 2013 00:58:56 +0200 Irek Szczesniak wrote:
> On Wed, Aug 14, 2013 at 9:37 PM, Glenn Fowler <[email protected]> wrote:
> >
> > On Wed, 14 Aug 2013 15:36:26 -0400 Glenn Fowler wrote:
> >> On Wed, 14 Aug 2013 21:29:50 +0200 Irek Szczesniak wrote:
> >> > On Wed, Aug 14, 2013 at 8:18 PM, Glenn Fowler <[email protected]>
> >> > wrote:
> >> > >
> >> > > invalid assumption that if this works
> >> > > open("/proc/self/fd/<DIRFD>/.")
> >> > > then other pathname calls will too
> >> > > and this is then used to emulate missing *at() calls
> >
> >> > Curse SUN and its half baked innovations.
> >> > if anyone of you have a patch send it to me or the list, please
> >
> >> yes
> >
> > yes to the curse
> How did you fix this for ast-ksh.2013-08-14? First I thought the code
> is clever and extracts the fd number from /proc/self/fd/<DIRFD>/ but
> then I found it just uses open() toopen /proc/self/fd/<DIRFD>/ and
> pass that to mkdirat(), which is a waste of good syscalls and doesn't
> work:
> ksh -c '/usr/bin/rm -Rf t1 ; mkdir t1 ; redirect {n}<t1 ; chmod a-r t1
> ; mkdir /proc/$$/fd/$n/t2 ; true'
> mkdir: /proc/645/fd/11/t2: [Permission denied]
> I would fix it myself but can't find the place which converts the
> /proc/$$/fd/$n/$path to a <fd>,<path> pair as it should be.
I've been sequestered with this code for the past few days and
am about to make a separate package just for testing
O_XATTR and /dev/fd/<FD/ and /proc/<PID>/fd/<FD/
the idea behind this work is to isolate all of the
#ifdef's errno some files maye be directories some directories may be files
dances into one place in libast
with that package I'll include a roadmap to the libast pieces involved
the hard part is the incomplete implementation of the kernel pieces in scope
both inter and intra vendor
_______________________________________________
ast-developers mailing list
[email protected]
http://lists.research.att.com/mailman/listinfo/ast-developers
